A project-wide search can be too successful. It may find a term in the current chapters, research notes, exported copies, and an archived draft at the same time. The total looks authoritative, yet it does not answer the practical question: how many matches belong to the folder being edited now?
This search selected folders text editor workflow treats search as an investigation of scope, not as a pretext for immediate replacement. It runs the same query once against the whole workspace and once against one recorded folder. The difference between those result sets reveals files that would otherwise contaminate the working count. No replacement command is part of this article.
Define the population before entering the query
The evidence workspace was /private/tmp/rune-g2c3-studio-321-_7t6eafa. It contained an included folder with three current files and an excluded folder containing archive.md. The query was Old321. Those three inputs—the workspace path, the selected-folder path, and the exact query—must be written down before evaluating any count.
An absolute path matters. A project can contain folders with similar names, a copied directory may move between sessions, and a remembered label is not proof of the scope passed to the command. Recording the path turns “I searched the current chapters” into a checkable statement.
Rune’s public documentation describes search at file, workspace, and selected file or folder scope. It also describes result counts and navigation from a result to its location. Those are documented product capabilities. The numeric results below come from a CLI run and should not be presented as a direct observation of the graphical interface.
Use the workspace result as a contamination baseline
The first command was rune search find --workspace /private/tmp/rune-g2c3-studio-321-_7t6eafa --query Old321 --json. Without a scope argument, it returned five matches in four files. One of those files was excluded/archive.md.
Five is not automatically the number of passages to edit. It is the workspace baseline. Its purpose is to show where the query occurs before the working population is narrowed. The baseline passes when all four file paths and five matches can be read and when the archived path is visibly part of that set.
A common failure is to copy the total into a task list and call every result current. Recovery begins by retaining the output, grouping paths into included and excluded locations, and running the scoped search as a separate operation. Do not silently delete the archive row from the baseline; its presence explains the difference that follows.
Run the selected-folder query independently
The second command added --scope /private/tmp/rune-g2c3-studio-321-_7t6eafa/included while preserving the same workspace and query. It returned four matches in three files. No result came from excluded/archive.md.
The scoped result passes only if both count and membership are correct. Three files and four matches are necessary, but every path must also sit below included, and the archive count must be zero. A total of four assembled from a wrong set of files would still fail. This path check prevents a coincidental number from being mistaken for proof of scope.
If an expected file is absent, verify the spelling of Old321, the file extension, and the folder path. If an unexpected file is present, inspect the scope argument rather than editing the result away. The query and path are independent inputs; change only one at a time so the reason for a new count stays visible.
This is the point where the investigation stops. Search results identify locations; they do not authorize changes. A passage may quote an obsolete name intentionally, document a migration, or belong to a reference file. Deciding what to rewrite is a different workflow with its own evidence and approval.
Recover differently from zero and excessive results
A zero result calls for input diagnosis. Copy the query again, inspect full-width and half-width characters, spaces, case, and punctuation, then verify that the selected folder actually contains the expected files. Preserve the zero-result input in the log. Jumping to workspace scope and accepting any result would hide the original mistake.
An excessive result calls for classification. Group the returned paths by their leading directory. Results outside the intended folder indicate a scope problem. Results entirely inside the folder may mean the query is too broad. If you refine the query, save that as a new run rather than overwriting the old one. The record must reveal whether the population changed or the search expression changed.
Avoid editing while assembling the comparison. A changed manuscript creates a new population and can invalidate the two counts. If edits have already occurred, repeat both searches and label the later pair as a new observation time.
For a graphical check, record the folder selected in the sidebar, the count displayed, and the paths reached by clicking representative results. That user-side check can confirm the interface behavior on the current installation. It should remain separate from the CLI transcript, which is the source of the four-file/five-match and three-file/four-match values here.
Close the search with an input-observation-decision record
A compact search record has three columns. Input contains the workspace, selected folder, and Old321. Observation contains four files/five matches for the workspace and three files/four matches for included. Decision states that one archived match belongs only to the workspace result and that the scoped result excludes it.
Acceptance requires the two commands to preserve the same query, the scoped set to contain only the three current files, and the archive to remain visible only in the baseline. Failure is any unexplained count or path. Recovery is a fresh, separately logged search after correcting either the scope or the query.
The search selected folders text editor method prevents two opposite mistakes: missing a current file and treating a reference file as current work. By ending the task after the scope is established, it also protects the archive from accidental modification. Search becomes a reliable map of the selected population, not an uncontrolled batch-edit button.